A Critical Role for the Right Fronto-Insular Cortex in Switching Between Central-Executive and Default-Mode Networks

A critical role for the right fronto-insular cortex in switching between central-executive and default-mode networks

Devarajan Sridharan*†‡, Daniel J. Levitin§, and Vinod Menon*†‡¶

*Department of Psychiatry and Behavioral Sciences, †Program in Neuroscience and ¶Neuroscience Institute at Stanford, Stanford University School of Medicine, Stanford, CA 94305 and §Department of Psychology, School of Computer Science and Program in Behavioural Neuroscience, McGill University, 1205 Avenue Penﬁeld, Montreal, QC, Canada H3A 1B1

Edited by Marcus E. Raichle, Washington University School of Medicine, St. Louis, MO, and approved June 20, 2008 (received for review January 1, 2008) Cognitively demanding tasks that evoke activation in the brain’s In a recent meta-analysis, Dosenbach and colleagues hypoth- central-executive network (CEN) have been consistently shown to esized that several brain regions that overlap with the CEN and evoke decreased activation (deactivation) in the default-mode SN are important for multiple cognitive control functions, network (DMN). The neural mechanisms underlying this switch including initiation, maintenance, and adjustment of attention between activation and deactivation of large-scale brain networks (7). However, no studies to date have directly assessed the remain completely unknown. Here, we use functional magnetic temporal dynamics and causal interactions of specific nodes resonance imaging (fMRI) to investigate the mechanisms underly- within the CEN, SN, and DMN. Converging evidence from a ing switching of brain networks in three different experiments. We number of brain imaging studies across several task domains first examined this switching process in an auditory event seg- suggests that the FIC and ACC nodes of the SN, in particular, mentation task. We observed significant activation of the CEN and respond to the degree of subjective salience, whether cognitive, deactivation of the DMN, along with activation of a third network homeostatic, or emotional (4, 8–11). The CEN, on the other comprising the right fronto-insular cortex (rFIC) and anterior cin- hand, is critical for the active maintenance and manipulation of gulate cortex (ACC), when participants perceived salient auditory information in working memory, and for judgment and decision event boundaries. Using chronometric techniques and Granger making in the context of goal directed behavior (12–18). We NEUROSCIENCE causality analysis, we show that the rFIC-ACC network, and the therefore hypothesized a key role for the SN in the hierarchical rFIC, in particular, plays a critical and causal role in switching initiation of cognitive control signals, specifically with respect to between the CEN and the DMN. We replicated this causal connec- activation and deactivation in the CEN and DMN, and the tivity pattern in two additional experiments: (i) a visual attention dynamics of switching between these two networks. ‘‘oddball’’ task and (ii) a task-free resting state. These results We used three functional magnetic resonance imaging (fMRI) indicate that the rFIC is likely to play a major role in switching experiments to examine the interaction between the SN, CEN, between distinct brain networks across task paradigms and stim- and DMN, with particular interest in the role of the FIC/ACC in ulus modalities. Our findings have important implications for a regulating these networks. In the first experiment, we scanned 18 unified view of network mechanisms underlying both exogenous participants as they listened with focused attention to classical and endogenous cognitive control. music symphonies inside the scanner. We analyzed brain responses during the occurrence of ‘‘movement transitions:’’ sa- brain networks ͉ cognitive control ͉ insula ͉ attention ͉ prefrontal cortex lient, orienting events arising from transitions between adjacent ‘‘movements’’ in the music (19). To specifically elucidate the role ne distinguishing feature of the human brain, compared of the FIC in driving network changes, we used chronometry and Owith brains lower on the phylogenetic ladder, is the amount Granger Causality Analysis (GCA), to provide information of cognitive control available for selecting, switching, and at- about the dynamics and directionality of signaling in cortical tending to salient events in the environment. Recent research circuits (20–22). suggests that the human brain is intrinsically organized into In the second experiment, we investigated the generality of distinct functional networks that support these processes (1–4). network switching mechanisms involving the FIC by examining Analysis of resting-state functional connectivity, using both brain responses elicited during a visual “oddball” attention task model-based and model-free approaches, has suggested the (23). A third experiment examined whether the network switch- existence of at least three canonical networks: (i) a central- ing mechanism could be observed during task-free resting state executive network (CEN), whose key nodes include the dorso- where there was no overt task and no behavioral response (4). lateral prefrontal cortex (DLPFC), and posterior parietal cortex Our motivation for examining the resting-state fMRI data was (PPC); (ii) the default-mode network (DMN), which includes the the recent finding, based on computer simulation of large-scale ventromedial prefrontal cortex (VMPFC) and posterior cingu- brain networks, that even in the absence of external stimuli, late cortex (PCC); and (iii) a salience network (SN), which certain nodes can regulate other nodes and function as hubs (24). includes the ventrolateral prefrontal cortex (VLPFC) and anterior insula (jointly referred to as the fronto-insular cortex; FIC) Author contributions: V.M. designed research; D.S., D.J.L., and V.M. performed research; and the anterior cingulate cortex (ACC) (1, 2, 4, 5). During the D.S. analyzed data; and D.S. and V.M. wrote the paper. performance of cognitively demanding tasks, the CEN and SN The authors declare no conflict of interest. typically show increases in activation whereas the DMN shows decreases in activation (1, 2, 6). However, what remains un- This article is a PNAS Direct Submission. known is the crucial issue of how the operation of these ‡To whom correspondence may be addressed at: Program in Neuroscience and Department of Psychiatry and Behavioral Sciences, Stanford University School of Medicine, 780 Welch networks, identified in the resting state, relate to their function Road, Room 201, Stanford, CA 94305-5778. E-mail: [email protected] or menon@ during cognitive information processing. Furthermore, the cog- stanford.edu. nitive control mechanisms that mediate concurrent activation This article contains supporting information online at www.pnas.org/cgi/content/full/ and deactivation within these large-scale brain networks during 0800005105/DCSupplemental. task performance are poorly understood. © 2008 by The National Academy of Sciences of the USA

www.pnas.org͞cgi͞doi͞10.1073͞pnas.0800005105 PNAS ͉ August 26, 2008 ͉ vol. 105 ͉ no. 34 ͉ 12569–12574 Downloaded by guest on October 1, 2021 Fig. 1. Activations in the Central-Executive and Salience Networks and deactivations in the Default-Mode Network during auditory event transitions. (A) Analysis with the General Linear Model (GLM) revealed regional activations (Left) in the right hemispheric FIC and ACC (blue circles); DLPFC and PPC (green circles) (coronal sections at y ϭϩ22, ϩ12 and Ϫ52 mm) and deactivations (Right) in the VMPFC and PCC (sagittal section at x ϭϩ4 mm and axial sections at z ϭϩ26 and Ϫ8 mm, yellow circles) during event transitions. The scale for t-scores is shown along side. Activations height and extent thresholded at the P Ͻ 0.01 level (corrected). (B) Independent Component Analysis (ICA, a model-free analysis technique) provided converging evidence for spatially independent and distinct networks. From left to right: Salience Network (rFIC and ACC), Central-Executive Network (rDLPFC and rPPC), and Default-Mode Network (VMPFC and PCC). Activations height and extent thresholded at the P Ͻ 0.001 level (uncorrected). The ICA prunes out extraneous activation and deactivation clusters visible in the GLM analysis to reveal brain regions that constitute independent and tightly coupled networks.

Our aim was to test the hypothesis that common network this method provides a way to estimate the peak latency of the switching mechanisms apply across tasks with varying cognitive BOLD response at each voxel using the ratio of the derivative to demands and differing stimulus modalities. If confirmed, our canonical parameter estimates (see SI Materials and Methods for findings would provide insights into fundamental control mech- details). This analysis revealed that the event-related fMRI anisms in the human brain. signal in the right FIC (rFIC) and ACC peaks earlier compared to the signal in the nodes of the CEN and DMN, indicating that Results the neural responses in the rFIC and ACC precede the CEN and We describe findings from Experiment 1 in the first three DMN (see Fig. S1 and Table S2). To provide converging sections. Convergent findings from Experiments 2 and 3 are quantitative evidence, we estimated the onset latency of the described subsequently. blood oxygen level dependent (BOLD) response in these regions using the method of Sterzer and Kleinschmidt (27). Previous Activation of CEN and SN, and Deactivation of DMN During Auditory studies have used differences in the onset latency of the BOLD Event Segmentation. As reported previously (19), we found robust response as a measure of the difference in onset of the under- right-lateralized activation in the DLPFC, PPC, and FIC during lying neural activity (20, 21, 27). We first defined regions of ‘‘movement transitions’’ in the auditory event segmentation task. interest (ROIs) in six key nodes of the SN, CEN, and DMN based Here, we extend these findings to characterize network-specific on the peaks of the ICA clusters (see Materials and Methods); all responses in the CEN, DMN, and SN. Activations in the CEN subsequent analyses was confined to these six canonical nodes of and SN were found to be accompanied by robust deactivation in the SN, CEN, and DMN (see also SI Text for a discussion on the the DMN at the movement transition [Fig. 1A and General choice of regions of interest and control analyses on regions not Linear Model Analysis in supporting information (SI) Materials included in the main analysis). We extracted the mean time- and Methods]. To further confirm that these regions constitute course in each of these six nodes, and used a sixth-order Fourier coherent networks, rather than isolated regional responses, we model to fit the event related BOLD response for each subject performed independent component analysis (ICA) on the task and event, and averaged the fitted responses across events and data, which revealed the existence of statistically independent subjects (see Fig. S2). Onset latencies were then computed as the CEN, SN, and DMN (Fig. 1B, see also Table S1) [ICA is a time point at which the slope of the fitted response reached 10% model-free analysis technique that produces a set of spatially of its maximum positive (or negative) slope in the initial ascend- independent components and associated time courses for each ing (or descending) segment. We found that the rFIC onsets subject (25)]. In the following two sections, we examine the significantly earlier than all of the nodes in the CEN and DMN putative causal mechanisms involved in switching between ac- (two-sample t-test, q Ͻ0.05; FDR correction for multiple com- tivation and deactivation in the context of the three networks, parisons) (Fig. 2, see also Table S3). These results confirm that identified above, using a combination of mental chronometry activity in the rFIC onsets earlier compared to the activation in and GCA (21, 22). the CEN nodes, and deactivation in the DMN nodes.

Latency Analysis Reveals Early Activation of the rFIC Relative to the GCA Reveals that the rFIC Is a Causal Outflow Hub at the Junction of CEN and DMN. First, we identified differences in the latency of the CEN and DMN. Finally, to elucidate the dynamic interactions the event-related fMRI responses across the entire brain using between the three networks we applied GCA. Briefly, GCA the method developed by Henson and colleagues (26). Briefly, detects causal interactions between brain regions by assessing the

12570 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0800005105 Sridharan et al. Downloaded by guest on October 1, 2021 null distributions of influence terms (F-values) and their differences (22). A causal connectivity graph was constructed using the thickness of connecting arrows to indicate the strengths of the causal influences (Fig. 3A, ‘‘raw’’ F-values normalized by the maximum F-value; raw F-values reported in Table S4). Only links that showed significant directed connectivity (influence terms) at the group-level (Mann-Whitney U test, P Ͻ 0.01; Bonferroni corrected for multiple comparisons) are shown (gray arrows, Fig. 3A); a subset of these links that showed a dominant direction of influence (difference of influence terms) are highlighted in red in the same figure (Mann-Whitney U test, P Ͻ 0.05, FDR corrected links shown in Table S4) (see SI Materials and Methods for details). GCA on the time-courses extracted from the key regions revealed statistically significant direct or indirect causal influences from the rFIC to all of the regions in the CEN and DMN (Fig. 3A). To quantify the causal interactions of each node of the network, we performed network analyses on key graph metrics (see Materials and Methods), and constructed a distri- Fig. 2. Onset latencies of the event-related responses in the six key nodes of the SN (blue bars), CEN (green bars) and DMN (yellow bars) in the auditory bution of these metrics, across subjects (for each node). Network event segmentation task. The rFIC onset signiﬁcantly earlier than each of the analysis on the causal flow network identified with GCA re- nodes in the CEN and DMN (two-sample t-test, q Ͻ0.05, indicated by (*), FDR vealed that the rFIC had the highest number of causal outflow corrected for multiple comparisons). Error bars denote standard error of the connections (out-degree), the lowest number of causal inflow mean (SEM) across subjects. connections (in-degree), and the shortest path length among all regions (means and standard errors of these metrics are reported in Table S5A). The rFIC also had a significantly higher net causal predictability of signal changes in one brain region based on the outflow (out-in degree) than all of the nodes of the CEN and time-course of responses in another brain region (28). We DMN (two-sample t test, P Ͻ 0.05). Differences in (out-in) performed GCA using a bivariate model (22) on the time- degree between the rFIC and the rDLPFC, rPPC, and PCC NEUROSCIENCE courses extracted from the six key regions used in the onset remained significant after FDR correction for multiple compar- latency analysis. We used bootstrap techniques (29) to create isons (q Ͻ 0.05) (Fig. 4A). Similarly, the rFIC had a significantly

Fig. 3. Granger causality analysis (GCA) of the six key nodes of the Salience (blue nodes), Central-Executive (green nodes) and Default-Mode (yellow nodes) networks during (A) auditory event segmentation, (B) visual oddball attention task, and (C) task-free resting state. GCA revealed significant causal outflow from the rFIC across tasks and stimulus modalities. In each subfigure, the thickness of the connecting arrows between two regions corresponds to the strength of directed connection (F-value) normalized by the maximum F-value between any pair of regions for that task (‘‘raw’’ F-values reported in Table S4). Only links that showed significant directed connectivity at the group-level (Mann-Whitney U test, P Ͻ 0.01; Bonferroni corrected for multiple comparisons) are shown (gray arrows); a subset of these links that showed a dominant directional influence (difference of influence term) are highlighted in red (Mann-Whitney U test, P Ͻ 0.05).

Sridharan et al. PNAS ͉ August 26, 2008 ͉ vol. 105 ͉ no. 34 ͉ 12571 Downloaded by guest on October 1, 2021 Fig. 4. Net Granger causal outflow (out-in degree) of the key nodes of the Salience, Central-Executive, and Default-Mode Networks in the three experiments. Comparison of the net causal outflow (out-in degree) for the six key nodes of the Salience, Central-Executive, and Default-Mode networks as assessed by Granger causality analysis revealed that the rFIC has a significantly higher net causal outflow than the CEN and DMN regions across tasks (conventions as in Fig. 2). Specifically, the rFIC had a significantly higher net causal outflow than almost all of the other CEN and DMN regions for the auditory segmentation and resting-state tasks, and the rDLPFC for the visual oddball task (two-sample t-test, q Ͻ 0.05, indicated by (*), FDR corrected for multiple comparisons).

shorter path length than all of the other regions except the exogenous and endogenous mechanisms underlying cognitive VMPFC (t test, P Ͻ 0.05); however, these differences did not control. remain significant after multiple comparison correction (data In the SI Discussion, we suggest that these interactions are the not shown). These results suggest that the rFIC is an outflow hub result of neural, rather than vascular processes. Here, we focus at the junction of the CEN and DMN. on the neurobiological implications of our findings in the context of the three networks that we set out to examine; analyses of Converging Evidence from Two Additional fMRI Experiments. To several other control regions (including the sensory and associ- provide converging evidence for the rFIC as a causal outflow ation cortices) that further clarify the crucial role of the FIC in hub, we analyzed fMRI data from two other experiments using the switching process are discussed in the SI Text. the same GCA and network analyses methods described above: (i) a visual ‘‘oddball’’ attention experiment, and (ii) a task-free FIC-ACC Network Is Neuroanatomically Uniquely Positioned to Gen- erate Control Signals. In primates, anatomical studies have re- resting state experiment (see also SI Materials and Methods). We vealed that the insular cortex is reciprocally connected to found a pattern of significant causal outflow from the rFIC that multiple sensory, motor, limbic, and association areas of the was strikingly similar to the auditory event segmentation exper- brain (30, 31). The FIC and ACC themselves share significant iment (Fig. 3 B and C). We then constructed network metrics for topographic reciprocal connectivity and form an anatomically these tasks using a procedure identical to the one used for the tightly coupled network ideally placed to integrate information auditory segmentation task. In each case, the rFIC had the from several brain regions (9, 10, 32). Indeed, analysis of the highest out-in degree and the shortest path length (Table S5 B auditory and visual experiments in our study found coactivation and C). Again, the rFIC again had a significantly higher net of these regions during task performance, as in many other causal outflow than several of the other nodes of the CEN and studies involving cognitively demanding tasks (7). Previous DMN (Fig. 4 B and C). Specifically, the rFIC had a significantly neurophysiological and brain imaging studies have shown that higher (out-in) degree than all of the other CEN and DMN the FIC-ACC complex moderates arousal during cognitively nodes in the resting state, and the rDLPFC in the visual oddball demanding tasks and that the rFIC, in particular, plays a critical task (two-sample t-test, q Ͻ 0.05, FDR correction for multiple role in the interoceptive awareness of both stimulus-induced and comparisons). These converging results indicate that that the stimulus-independent changes in homeostatic states (9, 10). rFIC is a critical, causal outflow hub across task paradigms and Furthermore, the FIC and ACC share a unique feature at the stimulus modalities. neuronal level: The human FIC-ACC network has a specialized class of neurons with distinctive anatomical and functional Discussion features that might facilitate the network switching process that ICA revealed the existence of statistically independent CEN, we report here. The von Economo neurons (VENs) are special- DMN, and SN during task performance, extending our recent ized neurons exclusively localized to the FIC and ACC (33). discovery of similar networks in task-free, resting-state, condi- Based on the dendritic architecture of the VENs, Allman and colleagues have proposed that ‘‘the function of the VENs may be tions (4). Our analysis indicates that the rFIC, a key node of the to provide a rapid relay to other parts of the brain of a simple SN, plays a critical and causal role in switching between the CEN signal derived from information processed within FI and ACC.’’ and the DMN (we use the term ‘‘causal’’ here, and in the (34). We propose that the VENs may, therefore, constitute the following sections in the sense implied by, and consistent with, neuronal basis of control signals generated by the FIC and ACC latency analysis, GCA and network analysis). The striking sim- in our study. Taken together, these findings suggest that the FIC ilarity of significant causal outflow from the rFIC across tasks, and ACC, anchored within the SN, are uniquely positioned to involving different stimulus modalities, indicates a general role initiate control signals that activate the CEN and deactivate for the rFIC in switching between two key brain networks. the DMN. Furthermore, our replication of this effect in the task-free resting state suggests that the rFIC is a network hub that can also Differential Roles of the rFIC, ACC, and Lateral Prefrontal Cortex in initiate spontaneous switching between the CEN and DMN (24). Initiating Control Signals. Many previous studies of attentional and Our findings help to provide a more unified perspective on cognitive control have reported coactivation of the FIC and

12572 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0800005105 Sridharan et al. Downloaded by guest on October 1, 2021 ACC (7, 23, 35, 36). The differential role of each of these regions regions within these networks during cognitive information has been poorly understood (37) as few studies have used processing—for understanding fundamental aspects of human chronometric techniques or causal analyses to dissociate the brain function and dysfunction. temporal and network dynamics of responses in these regions. We found that although onset latencies in the rFIC and ACC did Materials and Methods not differ significantly, as might be expected from their being Experimental Design. We used data from three different experiments. The ﬁrst part of the same (salience) network, the FIC did have a powerful experiment involved auditory event segmentation and detection of salient causal influence on the ACC (and correspondingly, higher net event boundaries in passages of music by the Baroque composer William causal outflow than the ACC) in all three datasets (Figs. 3 and Boyce. Eighteen right-handed participants (19–27 years of age, 8 females) 4). Even under conditions in which the ACC plays an important with little or no musical training participated in the experiment. Participants role in cognitive control (23, 36), the rFIC may generate the listened to stimuli with focused attention inside the scanner over noise- signals to trigger hierarchical control and previous studies, reducing headphones. A follow-up behavioral study conducted outside the scanner ensured that subjects could accurately detect the occurrence of including ours, may have missed detecting these effects. Our data movement transitions when these occurred in the stimulus. Further details can further suggest that when the ACC is dysfunctional (38, 39), the be found in Sridharan et al. (19). The second experiment involved a visual FIC is well positioned to trigger alternate cognitive control oddball task. Thirteen subjects (24 Ϯ 4.5 years of age, 8 females) participated mechanisms via the CEN. Our findings therefore help to clarify in the experiment. Two hundred visual stimuli were presented for 150 ms with an important controversy regarding the primacy and uniqueness a 2,000-ms interval between stimulus onsets. Visual stimuli consisted of col- of control signals in the prefrontal cortex (39). ored circles (either blue or green) and the frequency of the colored circles was Brain regions in the right inferior frontal cortex, surrounding counterbalanced across subjects (such that for half of them, green was the the FIC, have been implicated in a wide range of cognitive infrequent stimulus). Further details can be found in Crottaz-Herbette and control mechanisms (40–42). For example, many of the para- Menon (23). The third experiment involved an eight minute resting state scan digms involving response inhibition and inhibitory control have in which twenty-two subjects participated (19–21 years of age, 11 females). focused on ventrolateral regions (primarily within BA 47 and 45) Subjects were instructed to rest while keeping their eyes closed and were within the right inferior frontal gyrus (43). However, the specific requested to avoid moving during the scan (4). fMRI Acquisition, analysis with the General Linear Model (GLM), Indepen- role of the right FIC has been less well studied perhaps because dent Component Analysis (ICA), the Calculation of Peak and Onset Latency it is usually coactive with the lateral prefrontal cortex. A notable differences, and Granger Causality Analysis (GCA) followed the procedure exception is a study by Dosenbach et al. (44) who used resting- reported in a previously published experiment (19). Details can be found in the

state fMRI blocks, interspersed between task blocks, and graph SI. Here, we describe methods specifically related to network analysis of causal NEUROSCIENCE theoretical analysis to underscore the distinctiveness of the FIC interactions. and its connectivity with the ACC. Further, a recent lesion study in humans has shown that the rFIC has an important role in Region of Interest (ROI) Definition and Time Series Extraction. ROI analysis was cognitive control related to task switching. Using an oculomotor- performed using the Marsbar software package (http://marsbar.sourceforge. switching task Hodgson and colleagues (45) showed that patients net). Spherical ROIs were defined as the sets of voxels contained in 6–10-mm with lesions in the anterior rFIC were the most impaired in spheres centered on the peaks of activation clusters obtained from the ICA altering their behavior in accordance with the changing rules of analysis (Table S1). These same ROIs were used throughout all of the subse- the task. In normal healthy adults, functional brain imaging quent analyses (onset latency, GCA, and network analyses). The mean time course in each ROI was extracted by averaging the time courses of all of the studies have suggested that the FIC and the ACC are together voxels contained in the ROI. involved in a variety of cognitive control processes, including conflict and error monitoring, interference resolution, and Granger Causality Analysis. GCA was performed using the Causal Connectivity response selection (23, 36, 40, 46–48). We hypothesize that in all Toolbox (52), with modifications based on the methods proposed by Roebro- these cases, the rFIC enables task-related information process- eck et al. (22). GCA was performed on the timeseries extracted from ROIs to ing by initiating appropriate control signals to engage the ACC test for causal influences between ROIs taken pairwise using the difference of and the CEN. Our findings are inconsistent with the suggestion influence term (Fx3y Ϫ Fy3x) (22). We performed statistical inference on the that the FIC-ACC provides stable ‘set-maintenance’ over entire causal connections using bootstrap analysis: An empirical null distribution of task epochs whereas the fronto-parietal component initiates and the difference of influence terms was estimated using block-randomized time adjusts control (49). In our view, it is the FIC-ACC-centered SN series (22). For each subject, dominant (difference of influence) connections network that initiates key control signals in response to salient that passed a P ϭ 0.05 significance level (bootstrap threshold) were used for stimuli or events. As the lesion study by Hodgson and colleagues computing the network metrics described next. For details on the construction illustrates so dramatically, failure to generate these signals can of the causal connectivity graph (Fig. 3) refer to SI Materials and Methods. have severe consequences for behavior. Our findings do not, Network Analysis. To describe the interactions between brain regions in the however, preclude the possibility that after the FIC initiates causal network generated by GCA, we list the definition of the following changes in intra- and inter-network activity the CEN may carry metrics used in traditional graph-theoretic analyses (52): out top-down important control functions either on its own or in a) Out-degree: Number of causal outflow connections from a node in the association with the SN. network to any other node. Our findings help to synthesize these and other extant findings b) In-degree: Number of causal in-flow connections to a node in the in the literature into a common network dynamical framework network from any other node and they suggest a causal, and potentially critical, role for the c) (Out Ϫ In) degree: Difference between out-degree and in-degree is a rFIC in cognitive control. We propose that one fundamental measure of the net causal outflow from a node. mechanism underlying such control is a transient signal from the d) Path length: Shortest path from a node to every other node in the rFIC, which engages the brain’s attentional, working memory network (normalized by the number of nodes minus one). Shorter path and higher-order control processes while disengaging other lengths indicate a more strongly interconnected or ‘‘hub-like’’ node. systems that are not task-relevant. We predict that disruptions to For the present analysis, we constructed a distribution of these metrics, across subjects, for each node of the network. The mean value of these metrics these processes may constitute a key aspect of psychopathology (and their standard errors) across subjects are reported in Table S5. Path length in several neurological and psychiatric disorders, including fron- was computed using Dijkstra’s shortest path algorithm (53). A two-sample totemporal dementia, autism, and anxiety disorders (34, 50, 51). t-test was then applied on two key network metrics, the (out-in) degree and More generally, our study illustrates the power of a unified the path length, with FDR correction for multiple comparisons, to identify network approach—wherein we first specify intrinsic brain net- those nodes whose network metrics were significantly different from the works and then analyze interactions among anatomically discrete other nodes.

Sridharan et al. PNAS ͉ August 26, 2008 ͉ vol. 105 ͉ no. 34 ͉ 12573 Downloaded by guest on October 1, 2021 ACKNOWLEDGMENTS. We thank Mike Greicius for useful discussions and ford Graduate Fellowship to D.S. and by grants from the Natural Sciences and Elena Rykhlevskaia and Catie Chang for their comments on a preliminary draft Engineering Research Council of Canada to D.J.L., the National Science Foun- of this manuscript. We acknowledge two anonymous reviewers for their dation (BCS-0449927) to V.M. and D.J.L., and the National Institutes of Health insightful comments and suggestions. This research was supported by a Stan- (HD047520, NS058899) to V.M.

1. Greicius MD, Krasnow B, Reiss AL, Menon V (2003) Functional connectivity in the 27. Sterzer P, Kleinschmidt A (2007) A neural basis for inference in perceptual ambiguity. resting brain: A network analysis of the default mode hypothesis. Proc Natl Acad Sci Proc Natl Acad Sci USA 104:323–328. USA 100:253–258. 28. Abler B, et al. (2006) Investigating directed influences between activated brain areas 2. Fox MD, Corbetta M, Snyder AZ, Vincent JL, Raichle ME (2006) Spontaneous neuronal in a motor-response task using fMRI. Magn Reson Imaging 24:181–185. activity distinguishes human dorsal and ventral attention systems. Proc Natl Acad Sci 31. Efron B, Tibshirani RJ (1993) An introduction to the bootstrap (Monographs on USA 103:10046–10051. statistics and applied probability #57) (CRC Press LLC, Boca Raton, Florida) 3. Golland Y, et al. (2007) Extrinsic and intrinsic systems in the posterior cortex of the 30. Mufson EJ, Mesulam MM (1982) Insula of the old world monkey. II: Afferent cortical human brain revealed during natural sensory stimulation. Cereb Cortex 17:766–777. input and comments on the claustrum. J Comp Neurol 212:23–37. 4. Seeley WW, et al. (2007) Dissociable intrinsic connectivity networks for salience pro- 31. Mesulam MM, Mufson EJ (1982) Insula of the old world monkey. III: Efferent cortical cessing and executive control. J Neurosci 27:2349–2356. output and comments on function. J Comp Neurol 212:38–52. 5. Beckmann CF, DeLuca M, Devlin JT, Smith SM (2005) Investigations into resting-state 32. Mesulam MM (1998) From sensation to cognition. Brain 121(Pt 6):1013–1052. connectivity using independent component analysis. Philos Trans R Soc Lond B Biol Sci 33. Watson KK, Jones TK, Allman JM (2006) Dendritic architecture of the von economo 360:1001–1013. 6. Raichle ME, et al. (2001) A default mode of brain function. Proc Natl Acad Sci USA neurons. Neuroscience 141:1107–1112. 98:676–682. 34. Allman JM, Watson KK, Tetreault NA, Hakeem AY (2005) Intuition and autism: A 7. Dosenbach NU, et al. (2006) A core system for the implementation of task sets. Neuron possible role for von Economo neurons. Trends Cogn Sci 9:367–373. 50:799–812. 35. Fan J, McCandliss BD, Fossella J, Flombaum JI, Posner MI (2005) The activation of 8. Damasio AR (2000) The Feeling of What Happens: Body and Emotion in the Making attentional networks. Neuroimage 26:471–479. of Consciousness (Harcourt, Chicago) 36. Posner MI, Rothbart MK (2007) Research on attention networks as a model for the 9. Craig AD (2002) How do you feel? interoception: The sense of the physiological integration of psychological science. Annu Rev Psychol 58:1–23. condition of the body. Nat Rev Neurosci 3:655–666. 37. Milham MP, et al. (2001) The relative involvement of anterior cingulate and prefrontal 10. Critchley HD, Wiens S, Rotshtein P, Ohman A, Dolan RJ (2004) Neural systems support- cortex in attentional control depends on nature of conflict. Brain Res Cogn Brain Res ing interoceptive awareness. Nat Neurosci 7:189–195. 12:467–473. 11. Naqvi NH, Rudrauf D, Damasio H, Bechara A (2007) Damage to the insula disrupts 38. Baird A, et al. (2006) Cognitive functioning after medial frontal lobe damage including addiction to cigarette smoking. Science 315:531–534. the anterior cingulate cortex: A preliminary investigation. Brain Cogn 60:166–175. 12. Bunge SA, Ochsner KN, Desmond JE, Glover GH, Gabrieli JD (2001) Prefrontal regions 39. Fellows LK, Farah MJ (2005) Is anterior cingulate cortex necessary for cognitive control? involved in keeping information in and out of mind. Brain 124:2074–2086. Brain 128:788–796. 13. Crottaz-Herbette S, Anagnoson RT, Menon V (2004) Modality effects in verbal working 40. Nee TE, Wagner TD, Jonides J (2007) Interference resolution: Insights from a meta- memory: Differential prefrontal and parietal responses to auditory and visual stimuli. analysis of neuroimaging tasks. Cogn Affect Behav Neurosci 7:1–17. Neuroimage 21:340–351. 41. Aron AR, Fletcher PC, Bullmore ET, Sahakian BJ, Robbins TW (2003) Stop-signal inhi- 14. Petrides M (2005) Lateral prefrontal cortex: Architectonic and functional organization. bition disrupted by damage to right inferior frontal gyrus in humans. Nat Neurosci Philos Trans R Soc Lond B Biol Sci 360:781–795. 6:115–116. 15. Muller NG, Knight RT (2006) The functional neuroanatomy of working memory: Contributions of human brain lesion studies. Neuroscience 139:51–58. 42. Robbins TW (2007) Shifting and stopping: Fronto-striatal substrates, neurochemical 16. D’Esposito M (2007) From cognitive to neural models of working memory. Philos Trans modulation and clinical implications. Philos Trans R Soc Lond B Biol Sci 362:917– R Soc Lond B Biol Sci 362:761–772. 932. 17. Koechlin E, Summerfield C (2007) An information theoretical approach to prefrontal 43. Aron AR, Robbins TW, Poldrack RA (2004) Inhibition and the right inferior frontal executive function. Trends Cogn Sci 11:229–235. cortex. Trends Cogn Sci 8:170–177. 18. Miller EK Cohen JD (2001) An integrative theory of prefrontal cortex function. Annu 44. Dosenbach NU, et al. (2007) Distinct brain networks for adaptive and stable task control Rev Neurosci 24:167–202. in humans. Proc Natl Acad Sci USA 104:11073–11078. 19. Sridharan D, Levitin DJ, Chafe CH, Berger J, Menon V (2007) Neural dynamics of event 45. Hodgson T, et al. (2007) The role of the ventrolateral frontal cortex in inhibitory segmentation in music: Converging evidence for dissociable ventral and dorsal net- oculomotor control. Brain 130:1525–1537. works. Neuron 55:521–532. 46. Roberts KL, Hall DA (2008) Examining a supramodal network for conflict processing: A 20. Menon RS, Luknowsky DC, Gati JS (1998) Mental chronometry using latency-resolved systematic review and novel functional magnetic resonance imaging data for related functional MRI. Proc Natl Acad Sci USA 95:10902–10907. visual and auditory stroop tasks. J Cogn Neurosci 20:1063–1078. 21. Formisano E, Goebel R (2003) Tracking cognitive processes with functional MRI mental 47. Cole MW, Schneider W (2007) The cognitive control network: Integrated cortical chronometry. Curr Opin Neurobiol 13:174–181. regions with dissociable functions. Neuroimage 37:343–360. 22. Roebroeck A, Formisano E, Goebel R (2005) Mapping directed influence over the brain 48. Eichele T, et al. (2008) Prediction of human errors by maladaptive changes in event- using granger causality and fMRI. Neuroimage 25:230–242. related brain networks. Proc Natl Acad Sci USA 105:6173–6178. 23. Crottaz-Herbette S, Menon V (2006) Where and when the anterior cingulate cortex 49. Dosenbach NU, Fair DA, Cohen AL, Schlaggar BL, Petersen SE (2008) A dual-networks modulates attentional response: Combined fMRI and ERP evidence. J Cogn Neurosci architecture of top-down control. Trends Cogn Sci 12:99–105. 18:766–780. 50. Paulus MP (2007) Decision-making dysfunctions in psychiatry–altered homeostatic 24. Honey CJ, Kotter R, Breakspear M, Sporns O (2007) Network structure of cerebral cortex shapes functional connectivity on multiple time scales. Proc Natl Acad Sci USA processing? Science 318:602–606. 104:10240–10245. 51. Seeley WW, et al. (2008) Frontal paralimbic network atrophy in very mild behavioral 25. Beckmann CF, Smith SM (2004) Probablistic independent component analysis for variant frontotemporal dementia. Arch Neurol 65:249–255. functional magnetic resonance imaging. IEEE Trans Med Imaging 23:137–152. 52. Seth AK (2005) Causal connectivity of evolved neural networks during behavior. 26. Henson RN, Price CJ, Rugg MD, Turner R, Friston KJ (2002) Detecting latency differences Network Comput Neural Sys 16:35–54. in event-related BOLD responses: Application to words versus nonwords and initial 53. Dijkstra EW (1959) A note on two problems in connection with graphs. Numerische versus repeated face presentations. Neuroimage 15:83–97. Math 1:269–271.

12574 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0800005105 Sridharan et al. Downloaded by guest on October 1, 2021